On Edwin's point about changing minds, you have had an effect on changing my mind, although the preferred phrasing is of course, refined my thinking (if not quite in the way you may have intended). Observational physics naturally divides into the world outside of hadrons, and the inner hadron world. In the outer world there is space and time with associated inner particle symmetries spin (S3) and particle/anti-particle, isospin symmetry (S3) and hypercharge symmetry (S1), but no colour symmetry, giving an 8 dimensional outer world of most observations in physics (I think this was the point that Edwin was making).
In my case, this outer world is of a cyclical (S1) closed universe (S3) with electroweak base-space (S4) forming a product space. Here, gravity in S3 is by curvature and the particle forces of S4 by torsion, which are incompatible to be embedded in the same space. BUT in a local (vicinity) theory it may be possible to construct a torsion based view of gravity and so combine S3 and S4 into S7 to give your suggested Octonion Relativity covering the outer hadron world.
Joy's 2 particle correlation analysis for observables in the outer hadron world could be viewed as reaching the same conclusion. In the Standard Model, the electroweak charges are chiral and so spin orientation in S3 is related to isospin orientation in its S3, and hypercharge S1. The conclusion that correlations are due to the embedding of these spaces in S7 requires a flat parallelised sphere with torsion. There are 3 spaces here (physical space, symmetry space, measurement space) and whereas I deal with the first 2, Joy effectively deals with the second 2 - so direct comparison is a bit confusing. This is made worse in my GR model by compactified physical spaces being identified with symmetry spaces (as in Kaluza-Klein theories) - the physical scale factor between the 2 is the charge coupling constant.
In the inner hadron world there is also colour, which is not correlated with spin or the electroweak charges. This implies that Joy's correlation analysis for 2 colour observables (C, D) can be applied independently of the spin/electroweak observables (A, B) as there is no correlation between (A, B) and (C, D). The conclusion is the parallelised spheres S3 or S7, neither of which obviously squares with the Standard Model colour group SU(3). S7 isn't a group space, but S3 is - of SO(3). 2 particle correlation (the 2 quarks inside a meson) of an S3 colour space would yield S3 in Joy's analysis, whereas 3 particle correlation (a baryon) would yield the S7 result.
For your point of view, the inner hadron world yields a second independent occurrence of an S7 measurement space, as opposed to just the one in the outer hadron world. These inner and outer S7 occurrences could have the look of being mathematically dual, but as the references of Jens say that the E8 lattice is self-dual that may be a road to nowhere.
So you have persuaded me in principle, that in the outer hadron world it may be possible to construct a form of Octonion Relativity - which would necessarily be flat and not involve curvature - that could describe the measurement space of observables and particle symmetries. However, the spaces would not be physical spaces in the sense meant by Einstein. This would be apparent in particles having to be added by hand, and the theory not being extendable to the whole universe (that would be an invalid induction from a local vicinity theory). The corresponding physical theory for such an observational theory could of course be mine. I'm sticking with Albert on his concept of physicality and specifically a physical GR, as if I abandon this point then I would no longer consider myself a physicist.
There are 3 different types of space (physical, symmetry, measurement) and we may be confusing between them. We (me, you, Joy, Jens, ...) may not actually be handling separate beasts, but different parts of the same elephant.
Regards,
Michael